论文部分内容阅读
对众所周知的四种模拟语言加密处理方法,依照其剩余可懂度、带宽展宽和编码延迟三个方面进行了比较。可懂度得分是由约70个未受训练的收听者,对200个口述数字每个都给出识别回答所作的听觉试验并平均其正确数字识别的所得概率来得到的,此200个口述数字出现在平衡的50组四位加密数字发音表中。带宽展宽是以新的分段量度形式来表达的,对短时带宽运算来说它比常用的长时间平均功率谱量度更灵敏。编码延迟是模拟置乱器参数的一个直观函数。比较用的置乱方法是样点交换(S)、分组交换(B)、倒频(F)和B与F法的组合(BF)。样点交换包含有一个8千赫取样,其样点邻接的集合为L_S(2到128样点),而分组交换则对语言段邻接的集合N_B(4到128段)施行交换,其中每个段都是L_B(8到256)个样点长。倒频是通过每隔一个奈奎斯特(8千赫)样点简单地反一下符号来得到。L_S、N_B和L_B参数决定剩余可懂度及诸如延迟和带宽那样的传输特性。在我们的比较研究中对流行的BF制度给出一个合理的定量证明。例如,当N_B=8,L_B=128时,虽然编码延迟多达128毫秒,而带宽展宽仅约100周(采用新的分段量度法)(,数字可懂度Ⅰ为20%。需要指出:在识别十个数字的这种特殊问题中,完全随机(独立输入)时,收听者的回答其可懂度仅为I=10%
The well-known four kinds of simulation language encryption processing methods are compared according to their remaining intelligibility, bandwidth broadening and coding delay. The intelligibility score was obtained from about 70 untrained listeners, giving an auditory test of 200 answered dictations and averaging the probabilities of their correct number recognition, and the 200 oral figures Appears in the balance of 50 groups of four-digit encrypted digital pronunciation. Bandwidth broadening is expressed in terms of a new piecemetric measure, which is more sensitive to short-term bandwidth calculations than the commonly used long-term average power spectral measure. The coding delay is an intuitive function that simulates the scrambler parameters. The more scramble methods used are sample exchange (S), packet exchange (B), scrambling (F) and a combination of B and F methods (BF). The sample swapping consists of an 8 kHz sample with a sample contiguous set of L_S (2 to 128 samples) and a packet swapping of a set of contiguous segments N_B (4 to 128 segments), each of which The segments are all L_B (8 to 256) samples long. Scrambling is obtained by simply reversing the symbol every second Nyquist (8 kHz) sample. The L_S, N_B and L_B parameters determine the remaining intelligibility and transmission characteristics such as delay and bandwidth. In our comparative study, we give a reasonable quantitative proof of the popular BF system. For example, while N_B = 8 and L_B = 128, the code delay is as fast as 128 milliseconds and the bandwidth broadening is only about 100 weeks (using the new piecewise metric method) (the digitally intelligible degree I is 20%.) In this particular case of identifying ten digits, the listener’s answer is only I = 10% at a completely random (independent input)